Electricity is the moat, not the model

This week Microsoft withdrew from anchoring a data centre in Essex because the power would not arrive until the 2030s. In Asia, operators are ordering their own generation equipment. In San Francisco, the city paused new data-centre construction. None of these stories is about models. All of them are about electricity.

I run ELP GPT as an operating company, so I read these headlines as a procurement problem. The question is not which model is best this month; providers compete and our routing takes the cheapest one that meets the bar. The question is who can serve an answer at a known cost in 2029. That requires capacity you control, in a place with power you can count on.

That is why ELP Ventures has proposed investment in AI computing capacity in Ontario through 2030. Ontario has a low-carbon grid, a skilled workforce in Oakville and the GTA, and a regulatory climate that lets a company plan. We will finance that capacity against demand that already exists, not ahead of it, and we will report progress when it is real rather than when it is announced.

The honest version of the AI business is that the margin lives in operations: routing, reliability and power. We are building for that version.

— Edgar Melo Costa, founder

Founder’s notes are opinion. They are not investment, legal or professional advice.